Ex. 7.10 (TODO)

Ex. 7.10

Referring to the example in Section 7.10.3, suppose instead that all of the \(p\) predictors are binary, and hence there is no need to estimate split points. The predictors are independent of the class labels as before. Then if \(p\) is very large, we can probably find a predictor that splits the entire training data perfectly, and hence would split the validation data (one-fifth of data) perfectly as well. This predictor would therefore have zero cross-validation error. Does this mean that cross-validation does not provide a good estimate of test error in this situation? [This question was suggested by Li Ma.]

Warning

This problem has not been worked yet.